Photografting of Surface‐Assembled Hydrogel Prepolymers to Elastomeric Substrates for Production of Stimuli‐Responsive Microlens Arrays

Hydrogels have emerged as prototypical stimuli‐responsive materials with potential applications in soft robotics, microfluidics, tissue engineering, and adaptive optics. To leverage the full potential of these materials, fabrication techniques capable of simultaneous control of microstructure, device architecture, and interfacial stability, that is, adhesion of hydrogel components to support substrates, are needed. A universal strategy for the microfabrication of hydrogel‐based devices with robust substrate adhesion amenable to use in liquid environments would enable numerous applications. This manuscript reports a general approach for the facile production of covalently attached, ordered arrays of microscale hydrogels (microgels) on silicone supports. Specifically, silicone‐based templates are used to: i) drive mechanical assembly of prepolymer droplets into well‐defined geometries and morphologies, and ii) present appropriate conjugation moieties to fix gels in place during photoinitiated crosslinking via a “graft from” polymerization scheme. Automated processing enabled rapid microgel array production for characterization, testing, and application. Furthermore, the stimuli‐responsive microlensing properties of these arrays, via contractile modulated refractive index, are demonstrated. This process is directly applicable to the fabrication of adaptive optofluidic systems and can be further applied to advanced functional systems such as soft actuators and robotics, and 3D cell culture technologies.


Introduction
Hydrogels have emerged as prototypical stimuli-responsive materials with a wide range of applications in soft robotics, 3D cell Figure 1.a) Overview of the steps involved in the sample fabrication process.PDMS substrates are selectively oxidized through a MIMIC mask, then derivatized with a methacrylate-terminated silane.Array assembly proceeds by deposition of prepolymer solution onto the substrate under strain, followed by release of strain to promote collection of droplets into chemically wettable regions.Samples are cured by exposure to UV irradiation.b) Schematic of the biaxial stretch, assemble, set (BiSAS) device that executes the process.A syringe pump (SP) and a 3D-printed spray nozzle (b-i) are connected to a nitrogen gas line to nebulize and deposit prepolymer solutions onto the patterned PDMS surface assembling the hydrogel array (b-ii, scale bar is 500 μm).Three dynamic clamping arms and one anchor point are used to apply biaxial strain during deposition.c) Overview of surface derivatization and hydrogel network chemistry.A methacrylate-terminated silane enables photografting of PAm, PNIPAm, and PEG hydrogels via free radical polymerization initiated by LAP photo-initiator.
arrays and their application in robust optofluidic devices, where the focal length and magnification of the lenses are tuned by temperature or the solvent environment.
The procedure we report achieves three critical operations in one seamless fabrication scheme: i) Mechanically driven assembly of prepolymer solutions is executed using stretchable wettability patterns with well-defined and periodically arrayed hydrophilic zones with feature sizes ranging from 10's to 100's of μm; ii) Photoinitiated free radical polymerization is used to drive crosslinking of the hydrogel network, the chemistry of which is tuned to afford stimuli-specific response; and iii) Reactive moieties of the chemically designed wettability patterns result in photografting of the microgel structures to support substrate through a "graft from" polymerization process (Figure 1).This procedure enables the rational fabrication of large-area microgel arrays on elastomeric supports with robust hydrogel-substrate interfaces suitable for numerous applications in optofluidics, actuation, sensing, and cell culture.In this report, we focused on the fabrication and demonstration of dynamic optofluidic microlens arrays to highlight the liquid phase stability of the hydrogel-substrate interfaces throughout cycles of stimuli-induced contraction.
Microlens arrays have been applied in or suggested for a range of applications, including beam shaping, integral view photography, photolithography, optical computation, holography, and as biologically inspired optical sensors in soft robotics. [2]However, the widespread implementation of microlens arrays in these areas has been limited by the lack of facile, highly reproducible methods for the production of lens arrays with well controlled shape and size. [3]Moreover, current methods for the fabrication of microlens arrays often result in static lenses with fixed optical properties.
Hydrogels are excellent candidates for the assembly of dynamic microlens arrays because they undergo contraction or swelling in response to a wide range of environmental stimuli.Some examples of hydrogel stimuli include temperature, pH, solvent chemistry, ionic strength, electric fields, and even biological signals such as enzymes or antibodies. [4]Upon contraction or swelling, hydrogels can exhibit not only a change in refractive index associated with their water content but also a change in geometric features (e.g., curvature). [5]By tuning these critical parameters it is possible to control the optical properties of microgel lenses, which renders them amenable to, for example, optofluidic applications.In previous reports, microlens arrays have been prepared on hard substrates, where prepolymer solutions are cast into selectively etched wells, and on elastomeric substrates through the patterning of liquid droplets using surface functionalization of polydimethylsiloxane (PDMS). [6]The latter approach is advantageous because PDMS is readily derivatized through well-understood silane chemistry and enables the production of stretchable/flexible samples that can be strained or molded into 3D structures following fabrication. [7]Moreover, soft substrates such as PDMS can accommodate strain induced by morphological changes in gels at the interface, which supports their use as dynamic materials.
We previously demonstrated the use of protective masks to heterogeneously surface functionalize PDMS with defined patterns. [8]In combination with mechanical perturbation, this technique afforded the facile creation of large, well-ordered microdroplet arrays which behaved as microlenses.We have also shown the preparation of stimuli responsive arrays of hydrogels with a range of functional fillers that afford optical, magnetic, and humidity response. [9]However, these applications involved non-specific interactions between the droplet/gel and substrate surface, which resulted in weak adhesion that precluded fluidic applications.Herein we address this limitation by photografting microgels to a PDMS support for operation in different liquid environments.

Experimental Design
PDMS was chosen as the elastomeric support material because it was easily surface functionalized, optically transparent, and has a low Young's modulus.A chemical wettability pattern was generated through the application of micro-molding in capillaries (MIMIC) masks (the production of which has been described elsewhere) to the PDMS surface during oxidation (Figure 1a). [10]he use of MIMIC masks here allows for careful control of the size, shape, and periodicity of features in the resulting microgel array, but they were not required and other microgrids will also work.For example, similar oxidation patterns were easily generated using commercially available metallic microgrids or other periodic mesh materials (which do not require photolithography to generate) to mask the PDMS surface during plasma treatment. [8]Oxidation of PDMS enables surface derivatization with reactive double bonds by, for example, benzophenone and 3-(trimethoxysilyl)propyl methacrylate (TMSPMA) (S.2.2, Supporting Information).This allows subsequent photografting of hydrogel films via free radical polymerization chemistry through a "graft from" scheme. [11]Combining this chemistry with the use of MIMIC masks enables the selective functionalization of PDMS and generation of microgels with defined size, shape, and periodicity.Together, these techniques afford micro-structured hydrogel-elastomer hybrids with discrete geometry, which provides a number of advantages not offered by bulk bilayer filmsfirst, heterogeneous derivatization of the substrate enables discrete gel geometries, which confer improved stress management upon expansion or contraction (and so limits cracking, delamination, or other irreversible damage to the gel), fast mass/thermal equilibration, and control over micro/macro-morphology transitions when these materials were applied as actuators.Second, a discrete and periodic pattern of microgels affords their application as dynamic microlens arrays, which enabled the lensing demonstration shown here and expands the application space of such samples in the area of optofluidics.The Young's modulus of PDMS, which was in the range of 1.32-2.97MPa, enabled the convenient, low-force actuation of the support substrate to achieve mechanically-driven droplet assembly into an ordered array. [12]Sample fabrication was completed by exposure to UV radiation to induce photoinitiated polymerization and photografting of microgel networks to the substrate.This mechanism of polymerization was beneficial because the reaction does not occur before the array was assembled and proceeds rapidly at room temperature, which mitigates solvent evaporation from the prepolymer.This process also allows fabrication to be conducted with a single prepolymer solution, rather than a separate solution of monomer and free radical initiator.
A biaxial stretch, assemble, and set (BiSAS) device was designed and constructed to enable the automated deposition and mechanical assembly of prepolymer solutions into ordered arrays for photo-polymerization/grafting as described above (Figure S1 and S.2.3, Supporting Information).Although this device was not required for microgel assembly, it provides a convenient, automated platform for the application of biaxial strain to promote efficient and reproducible microdroplet assembly within hydrophilic regions of the surface template, which reduces offtarget structures.Specifically, biaxial strain of the substrate maximized pattern filling, hydrogel volume, and facilitated complete coalescence of droplets into patterned regions (Figure 1b).Uniaxial strain was previously used to promote droplet coalescence on selectively functionalized PDMS surfaces (Figure S2, Supporting Information). [8]Calculations showed that biaxial strain further increased the template surface area, enabling the accommodation of additional liquid during deposition, and thus making for a more efficient process (Figure S3, Supporting Information).A vertically mounted syringe pump and 3D-printed nozzle connected to a nitrogen gas line enabled the nebulization of prepolymer solutions for deposition onto the support material.A custom-designed co-axial air and fluid nozzle with a Luer lock attached to a 1 mL syringe loaded onto the syringe pump (Figure 1b-i) was used.The nozzle was 3D-printed through digital laser processing.Adjustment of the syringe pump rate, air pressure, and nozzle design allowed optimization of parameters for array assembly (Figure S4, Supporting Information).The inset of Figure 1b-i demonstrates the mist produced at 7 and 35 kPa air pressure, respectively.For all samples assembled with the BiSAS, the resulting spray was evaluated for droplet size and spray cone profile to provide maximum pattern filling.As an alternative approach for small batch production or the fabrication of microgels with small feature sizes (diameters below 100 μm), commercially available nebulizers were used for microdroplet assembly (S.2.5, Supporting Information).Following assembly samples were exposed to UV radiation by a portable UVM15 30,000 mW flashlight (365 nm) for 60 s yielding fixed microgel arrays (Figure 1b-ii).
Microgel arrays were fabricated both with and without the use of TMSPMA to investigate the advantages of photografting.Substrates prepared without TMSPMA cannot undergo radical polymerization with the hydrogel network and were expected to exhibit weaker microgel-substrate adhesion.Three hydrogel monomers were employed to demonstrate the versatility of the procedure: acrylamide (Am), N-isopropylacrylamide (NIPAm), and polyethylene glycol diacrylate (PEGDA 3400) (Figure 1c and see S.2.4,Supporting Information).These monomers were chosen because they were easily gelled via free radical polymerization (using chemical or photo initiators), compatible with aqueous prepolymer formulations, produce hydrogels that were sensitive to unique stimuli, and provide a prototypical hydrogel use set for various applications, including dynamic lensing.For example, others had shown polyacrylamide (PAm) hydrogels with solvent and pH response, poly(Nisopropylacrylamide) (PNIPAm) hydrogels with thermal or optical response, and polyethylene glycol (PEG)-protein hybrid gels which were sensitive to biological signaling molecules. [13]PNI-PAm and PEG microgel arrays were also broadly applicable in the areas of soft robotics and 3D tissue scaffolds, respectively. [14]ithium phenyl(2,4,6-trimethylbenzoyl)phosphinate (LAP) and N,N-methylenebisacrylamide were used for photoinitiation and as the Am and NIPAm crosslinker, respectively (Figure 1c).PEGDA served as both the monomer and crosslinker for the PEG microgels.
The morphology and geometric regularity of fabricated microgel arrays were characterized through a combination of optical and laser confocal microscopy, which provided information about the diameter, height, volume, and geometry of resulting products.Scanning electron microscopy (SEM) was employed to visualize the microgel-substrate interface and the dehydrated microstructure of the microgels.Microgel-substrate adhesion was characterized for samples prepared with and without photografting using combinations of bath ultrasonication, mechanical deformation, liquid shear stress in microfluidic channels, and tape adhesion tests.These tests represent a range of mechanical and environmental stresses on the microgel-substrate system which were expected to remove poorly adhered microgels.
The expanded application space and improved gel-substrate adhesion of the fabricated hydrogel arrays was demonstrated by investigating their capabilities as dynamic microlenses in liquid environments.Images printed on polyethylene terephthalate transparency sheets were projected through a PAm gel array, and solvent-induced changes in image magnification and focal length were observed using a microscope in transmission mode.The use of a microfluidic channel enabled the facile introduction of different solvent media, including water, ethanol, N,Ndimethylformamide (DMF) and concentrated aqueous sodium chloride.These solvents spanned a range of morphological states (swollen or contracted) of the microgels and demonstrated their dynamic and stimuli-responsive optical properties.The tunable stimuli-responsive nature of hydrogel microlens arrays were further demonstrated by subjecting PNIPAm gels, which undergo a thermally induced phase transition, to temperatures above and below the lower critical solution temperature (LCST).

Sample Preparation and Fidelity of Pattern Replication
The large-scale fabrication and implementation of microgel arrays requires an assembly method which can reliably produce samples with strong adhesion to a substrate, precise gel morphology, and minimal number of off-target structures.We have demonstrated previously that the latter two of these are well controlled through optimization of "master" patterns used to heterogeneously functionalize PDMS surfaces and application of strain to the substrate during assembly. [15]Briefly, the shape, packing arrangement, and resulting minimum edge-to-edge gap (Λ) between pattern features affected the incidence of particle deposition in hydrophobic interstitial regions (areas masked during oxidation) and particle bridging (the joining of two prepolymer droplets spanning an interstitial region).By straining and relaxing the substrate, it was possible to drive the assembly of interstitial droplets into hydrophilic regions.Small values of Λ were found to increase the incidence of particle bridging, while large Λ increased the population of interstitial droplets not absorbed into the hydrophilic zone.In general, Λ should be two to three times the diameter of the delivered prepolymer droplets to minimize bridging and interstitial microgels (e.g., the parameters utilized here delivered prepolymer droplets with average diameter of 16 ± 6 μm, yielding an optimal Λ in the range of 20-66 μm). [15]e employed the BiSAS to fabricate microgel arrays using both optimal (D = 300 μm, Λ = 50 μm; Figure 1b-ii) and suboptimal (D = 300 μm, Λ = 300 μm; Figure 2a) templates, observing quality ratios () of 0.51 ± 0.10 (n = 14) and 0.14 ± 0.05 (n = 84) respectively, defined at the ratio of target structures, N s , to total structures, T s .For the purposes of this study, we chose to fabricate microgel arrays using suboptimal templates with high Λ to contrast the adhesion of microgels assembled on surface functionalized and native PDMS-microgels formed on the hydrophobic interstitial regions of the PDMS template offered "built-in" controls ideally suited for direct comparison with microgels formed on the chemically modified zones of the PDMS template (see section 3.2).
We first assembled and crosslinked the three hydrogel materials (PEG, PAm, and PNIPAm) using equivalent templates (D = 300 μm, Λ = 300 μm), which give low  values.We used a combination of brightfield optical and laser confocal microscopy to characterize the geometric regularity and accuracy (vs the template), 3D structure (height and volume), and efficiency of fabrication.To facilitate image processing and analysis, we developed a MATLAB algorithm which rapidly output these parameters from raw image files enabling robust statistical comparisons (N > 8000 for each reported parameter collected from a minimum of 20 samples of each material) (Figure S5, Supporting Information).We collected composite optical/laser confocal images of samples to assess microgel array filling and diameter (Figure 2a) and topographical heat maps indicating microgel height to enable characterization of average gel volume (Figure 2b,c).
Generally, the assembled and crosslinked PEG, PAm, and PNI-PAm microgel arrays successfully replicate the template with few defects and statistically insignificant variations in diameter, height, and volume were observed for batches of the same gel type across each of the different materials systems.However, an ANOVA test revealed that there were statistically significant differences in the variance of morphological parameters between batches of gels of different types (see text S.3.3,Supporting Information), which we attribute to slight changes in prepolymer solution properties and equilibrium swelling ratios of the different gels.This analysis suggests that the fabrication process is reliably reproducible when using a single prepolymer solution, but that optimization will be required when switching between different prepolymer solutions.
To evaluate process efficiency, we focused on two parameters: 1) array occupancy %, defined as the percentage of available template sites occupied by a microgel (ideally 100%), and 2) interstitial %, defined as the percentage of the interstitial area occupied by an off target microgel (ideally 0%).We confirmed that the BiSaS can effectively process each prepolymer solution, as all three hydrogel systems exhibit statistically indifferent and high array occupancy %: 88 ± 9%, 90 ± 10%, and 90 ±10% for PEG, PAm, and PNIPAm, respectively (N > 8000 microgels, average 460 sites per sample) (Figure 2d).Most of the unfilled template sites (≈10% of those available) were toward the edge of the square assembly substrates.We attribute these vacancies to the fact that the aerosol stream produced by the BiSAS nozzle is cone shaped (Figure 1b-i), and ineffective at filling the corners of the substrate.Optimization of the substrate shape, nozzle stream, and/or the introduction of substrate rasterization would eliminate underfilling of the template.The interstitial % for this template (D = 300 μm, Λ = 300 μm) was 40 ± 17%, 31 ± 7%, and 25 ± 6%, for PEG, PAm, and PNIPAm, respectively (Figure 2d).We could easily decrease the amount of undesired interstitial product by utilizing optimal templates (Figure 1b-ii), but intentionally used such suboptimal patterns to facilitate internal controls for the adhesions tests that follow (see section 3.2.).
We employed confocal microscopy in air and liquid-phase optical sectioning microscopy to characterize the 3D morphology of dehydrated and hydrated microgels, respectively.Confocal microscopy allowed for the high throughput analysis of many gels (N > 100), but required the gels be dehydrated.In contrast, liquidphase optical sectioning enabled analysis of hydrated microgels, but this technique was relatively low throughput and microgels had to be imaged one at a time.We therefore adopted a combinatorial approach where we used ambient confocal microscopy to analyze large numbers of dehydrated microgels and liquid-phase optical sectioning to analyze a representative set of hydrated microgels.The complementarity of these two techniques allowed for a comprehensive understanding of micro gel morphology at the extremes of hydration/dehydration while balancing the constraints of sample throughput.For confocal measurements, we expected that, due to microgel contraction from dehydration, the diameter of the microgels would be less than that of the template (D = 300 μm).Indeed, we determined the microgel diameters in a dehydrated state to be 230 ± 30, 220 ± 30, and 210 ± 30 μm for PEG, PAm, and PNIPAm microgels, respectively, all of which are significantly smaller than the template diameter (Figure 2e and Figure S6, Supporting Information).These diameters were not significantly different while hydrated, measured as 230 ± 10, 210 ± 10, and 220 ± 20 μm for PEG, PAm, and PNIPAm, respectively (Figure S7, Supporting Information).We attribute differences between the microgel diameter and the template mask (observed in both hydration states) to: 1) the diffuse rather than sharp boundaries produced in the heterogeneous oxidation pattern at the contact point between the etch mask and substrate during plasma treatment; and 2) volume contraction of droplets during crosslinking owing to evaporation of liquid and evolution of dissolved gases imparted during droplet deposition.A representative image of a microdroplet array before and after curing by UV exposure demonstrates this volume contraction (Figure S8, Supporting Information).
Interestingly, although dehydrated PAm microgels exhibited a smaller diameter, they had a greater height (14 ± 5 μm) than PEG (10 ± 2 μm) or PNIPAm gels (9 ± 3 μm) (Figure 2f).This trend was not maintained when hydrated, as PEG hydrogels swelled to a height of 79 ± 5 μm, PAm to 65 ± 2 μm, and PNI-PAm to 36 ± 6 μm (Figure S7, Supporting Information).Differences in the heights of hydrated and dehydrated microgels are likely attributed to non-equivalent equilibrium swelling ratios for the hydrogel materials and morphological changes that each gel undergoes upon dehydration.We note that the microgels fabricated here possess nominal aspect ratios of ≈0.27 (calculated as the ratio of the maximum height to width in a hydrated state) ideally suited for dynamic lensing applications due to their thinlens-like morphology and fast mass transport/thermal equilibration times.For applications requiring higher aspect ratio microgels, further optimization of the nozzle design, spray conditions, or degree of strain of the support template can be employed; however, as the target aspect ratio is increased the stability of the structure will eventually become limited by the surface tension of the prepolymer, and sequential spray/set cycles or 3D templating procedures will be required.We note that confocal analyses for geometric parameters were carried out on samples following removal of interstitial microgels by ultrasonication (see section 3.2.for additional discussion) to minimize the computational burden in analyzing target structures.Spatial integration of microgel area with respect to height revealed average gel volumes of 0.6 ± 0.2, 0.7 ± 0.4, and 0.6 ± 0.3 nL for PEG, PAm, and PNIPAm gels, respectively (Figure 2g), which we measured to be statistically insignificant differences.

Characterization of Microgel-Substrate Adhesion
Strong microgel-substrate adhesion is critical to mechanically demanding and liquid phase applications.We characterized structure-substrate adhesion through a combination of microscopy and a series of adhesion tests.We first visualized the microgel-PDMS interface by imaging samples via SEM.Low magnification imaging of partial microgel arrays (Figure 3a) and higher magnification imaging of individual microgels (Figure 3b) revealed geometric regularities of the microgel arrays and differences in the microstructure of each microgel material.Specifically, surface roughness of the microgel materials was clearly different, with PEG and PNIPAm microgels being rougher than PAm microgels.These differences in roughness reflect differences in the microstructure of the microgels which are influenced by macromolecular parameters (e.g., crosslinking density, homogeneity, etc.), with PAm presenting the most idealized combination of features.Near the microgel-substrate boundary, PNI-PAm microgels exhibited increased wrinkling (Figure 3c), which may be indicative of inhomogeneous strain generation during contraction.Most importantly, we saw a clear difference between microgels produced on the chemically sensitized target regions and those produced on the non-sensitized interstitial regions (Figure 3d).Specifically, small protrusions "connecting" the microgel to the substrate were clearly visible for microgels produced in the target region where covalent crosslinking was expected (Figure 3c).In contrast, we observed a clear boundary between microgels and the substrate with no protrusions "connecting" to the PDMS template in the interstitial regions where crosslinking was not expected.We view the presence of protrusions in the target region (and their absence in the interstitial regions) as strong evidence of successful crosslinking of target microgels to the template.Interestingly, the overall microstructure of the PNIPAm microgels (not just the interfacial structure) was different depending on if they were in the target zones (attached) or in the interstitial zone (not attached) of the template.Specifically, PNIPAm microgels were rougher, with a more globular texture and greater apparent porosity if produced in the interstitial zone.This difference could be related to the tendency of PNIPAm to form microaggregates at temperatures above the LCST or in non-aqueous chemical environments, as others have demonstrated. [16]e characterized microgel adhesion to the target zones through a series of tests which compared templates treated with TMSPMA crosslinker and those that were not (oxidized PDMS).Microgels produced in the interstitial region of both template types (constituting native PDMS) should have equivalent, low adhesion, while those produced in the target zones should have different adhesion with the highest adhesion on TMSPMA treated templates (due to crosslinking).Following this logic, microgels in the interstitial region provided an internal control for each template, and as mentioned above, we intentionally used template geometries that gave low  and high populations of interstitial, poorly adhered microgels.We quantified adhesion as the percentage of microgels retained, calculated by enumerating the percentage of microgels which remain adhered to target zones after each test over the starting number of microgels.We employed three experiments to test microgel adhesion (Figure 4a and Figure S9, Supporting Information): a traditional tape peel test, a flow test inside a microfluidic device, and an ultrasonication bath test.We expected these tests to exert considerable stress on the substrate-microgel interface and remove any microgels not strongly adhered.Ultrasonication, for instance, is commonly used to clean surfaces, while tape tests are used to test thin film adhesion in many contexts. [17]The liquid flow tests provided insights into the performance of these materials in microfluidic and optofluidic applications where liquid shear forces are possible.Collectively, these tests constitute harsh conditions capable of removing poorly adhered microgels and provided tangible evidence of the benefits of photografting arrays to the assembly templates.
We investigated adhesion for PEG, PAm, and PNIPAm microgel arrays that were fabricated using equivalent templates (D = 300 μm, Λ = 300 μm) (Figure 4b,c).In all tests, we report a significant increase in microgel adhesion when the photografting molecule TMPSMA was used for surface crosslinking (Figures S10 and S11, Supporting Information).We performed a flow test by placing samples in a microfluidic channel (5 × 5 × 2 mm) and flowing water at rate of 10 mL per minute for 60 s.These conditions are particularly relevant for optofluidic applications such as the dynamic microlensing discussed below.The Reynolds number was 48 (where generally R e < 2000 constitutes laminar flow) and wall shear stress was 8.9 × 10 −3 dyn cm −2 (see S.3.2,Supporting Information).Under these conditions virtually all the microgels, regardless of the material, remained adhered to the TMSPMA treated templates (retention rates of 99.6 ± 0.3%, 98% ±1.5%, and 99.5 ± 0.6% for PEG, PAm, and PNI-PAm, respectively (Figure 4b,c).Further, the retention rates observed on TMSPSA treated PDMS were significantly higher than those on oxidized PDMS, where even the low sheer stresses of this test were sufficient to cause a majority of microgels to delaminate from oxidized PDMS within 1 min (Figure 4b).Surprisingly, the retention rate on oxidized PDMS for PAm (50 ± 30%) and PNIPAm (50 ± 30%) was much higher than PEG (2 ± 1.5%).To further examine this observation, we calculated the adhesive energies between oxidized PDMS and microgels in water using Equation ( 1) where E 132 is the adhesive energy of two materials across a medium (J m −2 ), A 132 is the Hamaker constant (J), and D is the equilibrium distance between two materials (≈0.2 nm) (see S.3.2,Supporting Information). [18]The calculated adhesive energies of −1.69, −2.55, and −1.22 J m −2 × 10 −3 for PEG, PAm, and PNIPAm, respectively, were very similar suggesting that electrostatics alone could not explain our observations.Specifically, Hamaker constants do not accurately reflect adhesion between materials with significant hydrogen bonding, and we believe the improved retention rates of PAm and PNIPAm on oxidized PDMS must therefore be explained by the tendency of amide and silanol groups to form hydrogen bonds at the microgel-PDMS interface. [19]Following the same argument, the poor retention rates of PEG (which contains no terminal alcohols) result in a lack of hydrogen bonding with the substrate. [20]For interstitial gels formed on native PDMS where hydrogen bonding does not contribute to adhesion, the retention rates for all microgel materials subjected to sonication or shear forces in the flow cell were low (<30%) and statistically indifferent (Figure S12, Supporting Information).We further investigated the impact of liquid flow on microgelsubstrate adhesion by monitoring an array of PNIPAm gels subjected to continuous, cyclical liquid flow (5.5 mL min −1 ) of hot (70 °C) and room temperature water at intervals of 1 min for a period totaling 4 h (Figure S13, Supporting Information).During this time, the microgels underwent 100 cycles of thermal contraction/expansion with none of the microgels monitored (N = 148) delaminating from the substrate.Together, these data confirm that photografted microgel arrays are stable and well suited for use in microfluidic applications.
We similarly quantified gel adhesion before and after subjecting samples to ultrasonication for 60 s.In this case, retention rates increased by 98 ± 1%, 86 ± 8%, and 80 ± 20%, for PEG, PAm, and PNIPAm samples, respectively, when the samples were treated with TMPSMA compared to an oxidized PDMS surface.Moreover, ultrasonication removed virtually all microgels in the interstitial regions of the pattern (Figure 4c and Figure S7, Supporting Information).As such, we could use ultrasonication as an additional processing step to remove interstitial microgels if needed.
Finally, we performed tape peel tests on dehydrated microgel arrays.Dehydration helped provide conformal contact between the tape and the microgels and it improved adhesion of the pressure-sensitive adhesive on the tape to the microgels.Again, photografting microgels to PDMS with TMSPMA increased retention rates over oxidized PDMS by 40 ± 40%, 50 ± 20%, and 74 ± 5% for PEG, PAm, and PNIPAm gels, respectively, but the differences were not as pronounced as in the previous ultrasonic and flow cell tests (Figure 4b and Figure S6, Supporting Information).Similarly, photografted microgels were retained over the interstitial microgels on native PDMS (Figure S14, Supporting Information), but the retention rates on native PDMS were highly variable.Using Hamaker constants calculated from literature values (see S.3.2,Supporting Information), we found the adhesive energy between hydrogels and tape to be −8.41,−7.52 2 , and −4.35 J m −2 × 10 −2 for PAm, PEG, and PNIPAm, respectively-values which are roughly double those expected for the adhesive energy between hydrogels and PDMS or oxidized PDMS (see S.3.2,Supporting Information).We therefore expected greater, statistically relevant differences in retention rates than what was observed.We attribute this discrepancy mainly to size differences in microgels (especially differences between interstitials and target gels), which make reproducible conformal contact of the tape impossible.Additionally, the dehydration process imparts heterogeneous stress at the microgel-substrate interfaces, further hindering reproducibility.For these reasons, we viewed the tape peel test results to be qualitatively representative of differences in adhesion between the microgels on different surfaces, while the ultrasonic and flow cell tests to be more indicative of the practical limits of adhesion for microgels crosslinked to the support substrate, especially within the context of the envisioned applications.
As a final test of the limits of photografted microgel adhesion, we subjected photografted microgel arrays to mechanical strain by stretching the underlying support substrate (e ≈ 1.3).Under these extreme conditions, photografted microgels remained robustly adhered to the support template, undergoing cohesive rather than adhesive failure (Figure S15, Supporting Information).Cohesive failure was also observed in the ultrasonication tests (Figure S16, Supporting Information).While we cannot directly measure the adhesive strength of the microgel-PDMS interface, this observation enabled an estimation of the lower bound of the adhesion strength, which we calculated to be to be 5 kPa (see S.3.1,Supporting Information).
Collectively, the adhesion test performed on PEG, PAm, and PNIPAm microgels demonstrate that derivatization of PDMS with TMSPMA to afford photografting of microgels significantly improves microgel-substrate attachment.This fabrication method produces robust samples which can withstand physical agitation, mechanical strain, and microfluidic flow in liquid environments.Although some sample delamination was measured even for the best performing system (PEG, with a minimum retention rate of 99 ± 1% across all tests), this degree of failure should not prohibit the application of samples in soft robotics, 3D cell culture, or optofluidic microlensing.Moreover, we could tune the chemical composition of prepolymer solutions to further improve adhesion.

Long-Range Array Order and Liquid Exposure
A constant and well-controlled periodicity is important for ordered microgel arrays to function effectively as microlens arrays.For dynamic functionality, this periodicity must be maintained upon morphological change of the microgel.We characterized the long-range order of the microgel arrays fabricated here by investigating the far field optical diffraction patterns that result from the constructive and destructive interference of diffracted light waves generated by an incident laser beam (Figure 5a). [21]pproximating the microgel array as a 2D array of slits, the principal angle of diffraction observed will be related to the spacing between slits according to the Bragg equation where d is the spacing between slits,  is the angle of diffraction, m (−1, 0, 1…) is the diffraction order, and  is the wavelength of the diffracted light.
To rapidly survey fluctuations in the long-range periodicity of the microgel arrays when swelled in different solvents, we measured the principal angle of diffraction of a standard laser probe, 650 nm (see S.2.9, Supporting Information).We used water, ethanol, DMF, and 5 m NaCl solutions as the submersion solvents because they induce volume changes to the microgels over a range of states of swelling/contraction.To easily visualize any differences in the resulting laser diffractograms, we prepared a microgel array with large periodicity (D = 70 μm, Λ = 30 μm) which increases .We used a 650 nm laser arranged perpendicular to a microgel array to produce a laser diffractogram which was projected on a screen 2.4 m away (Figure 5a).We collected the diffraction patterns with a camera placed 0.3 m behind the screen.The resulting diffractograms clearly demonstrate the long-range order of the microgel arrays (Figure 5b and Figure S17, Supporting Information).We measured the scattering vector, ⃗ g, to quantitatively compare diffractograms produced by microgel arrays in the different solvents.Specifically, we measured ⃗ g and calculated the principal diffraction angle using the relation: L where L is the distance between the array and the screen (Figure 5b).In tandem we measured the change in the topdown area of the microgels for the same solvents using brightfield microscopy (Figure 5c; water, 5 m NaCl, ethanol, and DMF).These analyses confirmed that despite significant changes to the volume of the microgels through solvent swelling/contraction, a well-defined diffractogram was always observed and there were no statistically significant difference between the diffraction angles produced (Figure 5c and see S.3.3,Supporting Information).We can therefore confirm that the overall periodicity and geometric order of the microgel arrays are maintained throughout significant solvent swelling/contraction cycles (≈400 microgels, 0.04 cm 2 , were illuminated during diffraction).Calculation of d in Equation (2) from the four diffractograms gives an average periodicity of 102 ± 0.1 μm, which was slightly larger than the 100 μm periodicity of the chemical template.We attributed this difference to the slight strain on the PDMS as it is laminated to the glass slide.The robust microgel-substrate adhesion afforded by our fabrication process supports dynamic changes in volume to the individual microgels without compromising the overall fidelity of the array.

Environmentally Responsive Dynamic Microlens Arrays
To demonstrate the optofluidic capabilities of fabricated microgel arrays, we loaded samples into a microfluidic test cell comprised of 3D-printed supports, pressure-sealed glass slides, and a molded PDMS microfluidic channel (Figure 6a).The device could accommodate samples up to 5 × 5 mm in area and solvents are easily exchanged using standard tubing and pumping systems.The small radius of curvature of the microgels resulted in small focal lengths with imaging performance that is most readily studied using transmission optical microscopy.In a prototypical experiment, we used an LED light to illuminate an object beneath the microgel array (contained in the microfluidic test cell) that was subsequently focused into an array of microscopic images by the microgel lenses which were collected simultaneously using an upright microscope (Figure 6b and Figure S18, Supporting Information).
The lensing properties of a microgel are described by the lens maker's formula where f is the focal length, Δn is the difference in refractive index between the lens n l and the surrounding medium n 0 , and R 1 and R 2 are the incident and exterior radii of curvature, respectively. [21] ray diagram illustrating the lensing activity of a microgel includes the object (u) and image (v) distances as related by the focal length and object size (h) and image size (h′) (Figure 6c).In this experiment, we positioned the elastomeric microgel array on the interior side of the top glass plate of the microfluidic device.As a result, R 1 corresponds to the curved surface of the hydrogel, while R 2 corresponds to the flat side which was adhered to the substrate.The contribution of 1 R 2 in Equation ( 3) was negligible, because the radius of curvature of a flat surface is infinite and 1 R 2 approaches zero.We expected the contraction and expansion of hydrogel materials to alter the focal length and magnification of the microgel lens yielding observable changes to the projected image.Object and image distance are related to focal length by the Gaussian form for thin lenses, Equation (4), while magnification, M, can be determined as a function of projected image size or object and image distance, Equation (5). [22] According to Equations ( 3)-( 5), the lensing properties of a microgel can be altered either by a change in Δn n 0 or R 1 , which causes focal length f and magnification M to change.We explored whether the radius of curvature R 1 dictates the lensing properties of a gel by collecting cross-sections of fluorescent Zstack images of a polyacrylamide hydrogel in water and ethanol (Figure 6d).The image collected in water represented the fully hydrated state of the microgel, while the hygroscopic nature of ethanol decreased the water content of the gel and induced roughly a 40% contraction.In water, we measured the radius of curvature to be 30 μm, compared to an increased value of 32 μm in ethanol.Through analysis of Equations ( 3)-( 5) and by assuming that the image distance v was much less than the object distance u in this experiment, S.3.5, Supporting Information, the proportion can be derived.This relationship indicates that, if the radius of curvature primarily dictates a microgel's lensing properties, an increase in R 1 such as for ethanol would increase the size of the image (increase magnification relative to water).However, from our tests, ethanol caused the greatest contraction in PAm microgels and produced the most demagnified image of all solvents tested (M = 0.0022) (Figure 6e).We observed image demagnification values of 0.0039 (concentrated NaCl), 0.0033 (water), and 0.0031 (DMF) (Figure 6e).
According to these results, the relation between M and R 1 cannot be purely a function of R 1 .Instead, the lensing properties of the microgels must also result from changes to the term Δn n 0 from Equation (3).In this expression n 0 is equal to that of the ambient environment, which will depend purely on the solvent used, thus n 0 is equal to the refractive index of 1.33, 1.36, 1.38, and 1.43 for water, ethanol, 5 m NaCl, and DMF, respectively. [23]Concerning n l , others have reported that the refractive index of a hydrogel changes with water content during contraction or expansion. [24]pecifically, contracted hydrogels have a greater optical density, and contraction of a hydrogel microlens is understood to alter its focal length. [25]We can therefore expect the term dominating changes in optical properties of the microgel lenses to be Δn n 0 (see S.3.5, Supporting Information), assigning the following proportionality For example, a solution of 5 m NaCl (with increased n 0 compared to water) did not significantly change the volume of a PAm microgel, which results in a smaller term for Δn n 0 and consequently lowers demagnification.DMF both significantly increased n 0 and caused the gel to contract, which resulted in an increase in both terms.Correspondingly, the change in demagnification is negligible compared to water, which would be consistent with the change in n 0 and n l effectively negating one another.Finally, submersion of microgels in ethanol induced substantial contraction but only slightly changed n o compared to water.As such, the projected image was most demagnified for ethanol.We further explored the relationship between magnification and Δn n 0 by examining the demagnification afforded by dehydrated microgels in air, where the terms Δn and n 0 are the greatest and smallest, respectively (Figure S19, Supporting Information).In agreement with the described trend, the projected image was significantly more demagnified in air than for any of the solvents tested.
Although PAm microgels limited the dynamic microlensing capabilities of the array to solvent response, the chemistry of the hydrogel network can be easily tuned to afford alternative stimulus sensitivity.For example, we fabricated a microlens array using PNIPAm, which exhibits thermally responsive properties due to its LCST.This material enabled the facile alteration of microgel optical properties by changing the temperature.
Thermal cycling is advantageous, because it can be activated globally by heating the entire device or locally by targeted exothermic reactions or selective near-infrared irradiation and it does not require exposure to harsh solvents or solutions. [26]We demonstrated thermal cycling of a PNIPAm microlens arrays by exposure of microgels to water at temperatures above and below the LCST of PNIPAm using the microfluidic test system from above (Figure 7a and Video S2, Supporting Information).Specifically, at temperatures below the LCST, the projected image was in focus and the hydrogel was fully swollen.When hot water (≈80 °C) was introduced to the flow cell, the hydrogel contracted, and the focal plane of the image shifted out of focus (Figure 7b and Video S3, Supporting Information).In this experiment, we maintained a constant distance between the projected image and camera, so changes in image focus were affected only by the variable focal length of the microgel.
We observed that the refractive index of microgels within an array had the greatest impact on their optical properties.PNI-PAm gels at temperatures above the LCST (in a contracted state) exhibited increased image demagnification (Figure S20, Supporting Information), which is explained by an increase in the refractive index of a microgel as a function of water content.Although the refractive index of water, n 0 , is known to depend on temperature and decreases as temperature increases, the comparatively larger increase in n l counteracts this effect. [27]The optical properties of PNIPAm microgels can be cycled repeatedly by exposure to water above and below the LCST of NIPAm (Figure 7c) and could be employed to produce variable focus microgel lens arrays.We observed a statistically significant difference in the overall average magnification of images projected through microgels in a contracted (0.0043 ± 0.0000 1 ) or expanded (0.0054 ± 0.0000 1 ) state over 100 cycles between 70 °C and room temperature water (N > 140 projected images).Representative images of the focal plane of the projected object clearly demonstrate the difference in magnification when gels are in a contracted or expanded state (Figure S21, Supporting Information).The increased variability in magnification for microgels in the swollen state (cold water) is explained by an apparent increase in the depth of field (Figure S22, Supporting Information), which complicated determination of the exact focal plane of the projected image.We expect that incorporation of software which determines when projected images are in focus would significantly reduce variability and provide more precise magnification.However, we note that each cycle shows a clear and statistically significant difference in magnification (averaging 28 ± 7% additional demagnification in a contracted state) for PNIPAm gels in water above or below the LCST.Collectively, these results demonstrate the ability of the reported fabrication method to produce liquid-stable stimuli-responsive dynamic microlens arrays with robust adhesion to an elastomeric template.We expect these results to support the broad application of specifically tuned microgel chemistries as dynamic gel arrays for use in areas such as soft robotics, optofluidics, and sensing.

Conclusions
We have presented a facile and versatile method for the microfabrication of liquid-stable, elastomer-attached, large-area hydrogel arrays of controlled size, shape, periodicity, and chemical composition.This procedure involves three critical operations: i) heterogeneous oxidation of an elastomeric substrate with a physical mask and mechanically driven assembly of prepolymer droplets within a defined pattern; ii) surface functionalization of oxidized zones with reactive moieties (e.g., methacrylate groups) to drive the covalent attachment of microgels to the elastomeric template; iii) photoinitiated polymerization of the assembled array to drive crosslinking of hydrogels with tunable chemistry and stimulus response.We demonstrated the fabrication of well-ordered features with diameters from tens to hundreds of μm using PAm, PNIPAm, and PEG hydrogels.In contrast to previous methods for the microfabrication of gel arrays on elastomeric sub-strates, this technique provides robustly adhered, liquid-stable microgel arrays with tunable stimulus response.We characterized the adhesion of microgels to PDMS templates when subjected to ultrasonication, shear forces in microfluidic channels, mechanical strain, and tape peel tests.The use of a graft-from fabrication scheme significantly improved microgel-substrate adhesion, which rendered fabricated arrays amenable to liquidsubmerged applications without sacrificing the morphological stimulus responsive nature of the microgels.We demonstrated the high degree of array ordering maintained when samples were exposed to solvents which induced up to a 40% reduction in microgel area.Finally, we prepared dynamic microlens arrays with solvo-thermal response using PAm and PNIPAm microgels as model systems.The microgels demonstrated stimuli-tunable, dynamic microlensing characteristics in simple optofluidic devices.These microgel arrays are broadly applicable to the fabrication of stimuli-responsive microstructures applicable to soft robotics, 3D culture, and optofluidics.For example, specifically designed masks could be used to pattern morphologically controlled microgels on substrates to afford programmable actuation modes in response to chemical stimuli or in ways that mimic biological systems.Likewise, incorporation of cells into the microgels allows for simultaneous culturing of hundreds of replicants in one sample and the variable size of the hydrogels can be extended to more complex biological components such as organoids.Furthermore, the chemistry of the prepolymer mixture can be easily altered to afford a wide range of mechanical properties or environmental response beyond those listed.

Figure 4 .
Figure 4. a) Illustrations of the tests performed to quantify adhesion for microgel arrays.b) Percentage of microgel arrays retained on average for each hydrogel with and without photografting.c) Optical microscopy images of a representative portion of PEG hydrogel arrays before and after each test.Scale bars are 500 μm.

Figure 5 .
Figure 5. a) Experimental setup employed to measure the laser diffractograms produced by a microgel array in different liquid media.b) Example diffraction pattern.Vector ⃗ g represents the measured scattering vector.Image was color inverted to increase contrast for analysis.c) Diffraction angle and hydrogel area in different liquid media.

Figure 6 .
Figure 6.a) Schematic of the flow cell used to exchange submersion fluid for the microgel lens arrays.b) Schematic of the optofluidic lensing experiment.A photograph of the actual flow cell device is provided as inset.c) Ray diagram of a single microgel acting as a microlens.Object size h and image size h′ are used to determine magnification.d) Cross-section of Z-stack images of a microgel in water and ethanol.e) (Top row) Optical micrographs of a PAm microgel in NaCl, water, DMF, and EtOH (annotated dotted red circle provides microgel diameter in water), and (Bottom row) optical micrographs of the projected letter "N" (annotation provides magnification).

Figure 7 .
Figure 7. a) Optical micrographs of thermally controlled microlensing.At the start (t = 0 s), temperature is below the LCST of the PNIPAm microgel (T = 7 °C) and the projected letter "N" is in focus.Proceeding clockwise from t = 0 s to t = 100 s, temperature was raised above (T = 80 °C) and returned to below the LCST of PNIPAm, switching between out of focus and focused "N".b) Corresponding optical micrographs of a single microgel during the process in panel (a).c) Magnification of the projected images in cold and hot water.Scale bars are 50 μm.Average magnification in each environment is highlighted by a horizontal dashed line.Gray shading indicates ± 1 standard deviation (N > 140 projected images for each cycle).